/HexMCTS.jl

Monte Carlo Tree Search to play Hexgonal WuZi(Five in a Row), including MCTS algorithm backend in julia and GameUI in python

Primary LanguageJuliaMIT LicenseMIT

HexMCTS.jl

Intro

This is a julia implementation of HexWuZi, including MCTS algorithm in julia and GameClient in python.

This julia version is much faster than python with numba version, see Performance comparison of each version section for more detail.

What's new!

Since the old version can't run well on my new macbook, I write a new julia program to host a web server to communicate between julia backend and python fontend instead. This new version can run well both on macos and windows.

Play it

  1. Start the BackendServer with running the BackendServer.jl

    julia BackendServer.jl

  2. Start the GameClient with running the GameClient.py

    python GameClient.py

However, you should install all dependency first, please follow Requirement section to install them well.

Requirement

Julia

Go to Juila Official Website to download and install one.

Oxygen

We use Oxygen package for web development in julia, you have to install it first.

import Pkg
Pkg.add("Oxygen")

Since the old version with julia package in python and PyCall.jl in julia can't run well on my new macbook, I write a new julia program to host a web server to communicate between julia backend and python fontend instead.

Performance comparison of each version

Version Perfomance(times of simulation in 5 seconds)
"Pure Python" 800-900
Python with Numba 8k-9k
Pure Julia 35k-40k

Why there is quotation mark around "Pure Python"?

Because if you use NumPy, then it's not pure any more since NumPy is written in C.